As the above quotation by Paul Samuelson suggests, textbooks play a central role in economics. They shape the framework within which economists look at issues, determine which questions are posed, and play a role in deciding who chooses to become an economist. Despite their importance, textbooks have received little attention from economists.

Principles textbooks do not have a good image among serious economists; they generally believe that texts reduce economists' profound thoughts into oversimplified models and maxims that reflect yesterday's ideas. Intermediate texts are seen as more closely following modern ideas, but even these oversimplify and follow new developments with a long and variable lag.

I agree with many of the criticisms of texts, and have made many of them myself. But, once I became a textbook author, I discovered the constraints under which textbook writers operate. These constraints push successful texts to their simple mantras and litanies. In large part, the degree to which my texts are successful reflects my willingness to compromise, and, to some degree, the failure of my texts to do even better than they do reflects my unwillingness to compromise–my desire to use texts to allow and encourage a somewhat different treatment of the canon than presently exists. Because of these constraints Samuelson's image of the powerful, controlling textbook writers is incomplete. Textbook writers are controlled as much as they are controlling.

To understand that control, one must understand the process by which textbooks are produced. Textbooks are money-making propositions, published by large corporations whose focus is on the bottom line. To see that textbooks will make money, all textbooks are heavily reviewed; we're talking 50 to 100 reviews for a principles text. These reviews are not for accuracy, nor are they to see that the ideas in the books are up to date, or even correct; the reviews exist to find out if those professors who teach the course are comfortable with the approach the author has taken. It is for that reason that texts usually reflect a merger of ideas that flourished over the past 30 years, when the now current professors went to graduate school, with a smattering of current ideas that have become popular and well known.

All successful textbook authors are responsive to these reviews. If one wants to be a player one must cater to the large portion of the market, making the book seem sufficiently traditional so that the majority of professors find it easy to teach from, using their old notes, but sufficiently new to give them a feeling that the book is "modern." This is the reality: Successful textbooks are those that are adopted by a large group of professors, the majority of whom often are removed from the front lines of economic theory, and some of whom have had little training in economics. (In some community colleges, anyone who has been trained in business or social science is seen as an acceptable teacher of the principles of economics.) Because of the reviewing process a type of textbook lock-in occurs. Textbooks acquire a life of their own–a life which no one individual controls. If one wants to introduce new ideas into the texts one must take this textbook lock-in into account.

In this paper I consider the Santa Fe work on complexity in relation to economics textbooks, asking the question: How can the Santa Fe work best be integrated into the texts? In my view the speed with which economics texts integrate Santa Fe ideas will play an important role in determining whether complexity remains an interest of a small group of the profession or whether it becomes the foundation of a revolution in economic thinking that I think it should be.

The Santa Fe approach is only in its infancy, but for those of us bitten by the Santa Fe bug, it offers a path to a much more meaningful economic theory. However, if it is to blossom, we need to increase the number of individuals open to the Santa Fe "process" way of looking at the economy. Current texts are slowing its advance in two ways. First, they are weeding out those students most likely to look at economics within a process view, since such students find the current texts discordant with their understanding, and choose not to study economics. Second, they are setting a static vision for those who stay within economics, making it difficult for them to see the economy within a Santa Fe framework. The models and conceptual pictures presented in the current texts become mantras that are repeated again and again, and, as they are, these models and conceptual pictures are no longer caricatures of economic thinking, but the foundation of economic thinking itself. Thus, textbooks direct the mindset of future economists and, in doing so, influence the research the profession does.

Consider some of the reactions to my attempting to include a Santa Fe view of growth into an intermediate macro text. Some of the comments the reviewers wrote were:

Because of these reviews, if my intermediate macro book is to be published, I will have to relegate the Santa Fe view of growth to a box. My view of integrating the Santa Fe approach into textbooks has been significantly shaped by experiences such as these.

There are essentially two ways in which the Santa Fe vision can be introduced into economics texts. One way is to develop a completely new Santa Fe style text. Such a text would be fundamentally different from current texts; it would logically develop the central ideas of iterative dynamic processes, introduce students to topics in the science of complexity, such as Mandlebrot and Julia sets, and then draw implications from these about economics. Such a text would find a niche, and, eventually, that niche might expand. The problem with that approach is the textbook lock-in I spoke of earlier. The initial niche would be small, and it would grow slowly, if at all. .Given the inertia in the teaching profession, I see a second way as being more likely to hasten the spread of the Santa Fe approach. That second way is to kneed the Santa Fe ideas into simple mantras and conceptual pictures that are sufficiently compatible with existing models and conceptual pictures so that those ideas are integratable into existing texts. It is that evolutionary approach that I discuss in this paper.

The remainder of the paper is organized as follows: First I briefly discuss the content and vision of current textbooks, together with their core models. Next I discuss the Santa Fe vision and how it differs from the current vision in the texts. Finally, I discuss three general Santa Fe types of ideas that I believe can be added to the texts (if done discreetly) and suggest how those ideas can be integrated into the standard models that form the core of principles texts.

The vision behind almost all current texts is the neoclassical vision. This vision has, at its core, a conception of a deductively-determined equilibrium–the perfectly competitive state towards which economic forces drive the economy. It is a vision that requires the arrow of time to be downplayed, and the deductive analytics to be emphasized. It is a vision that emphasizes the reducibility of knowledge–one that sees representative agents as relevant, and one which essentially sees the interaction of the economy as reducible to a set of static models that are analytically solvable. Thus, if one had a super computer, and sufficient computing time, one could "solve" the economic problem, and choose the "right" prices.

In the texts, this vision is reduced to the following mantra: The world is full of tradeoffs; markets guide people in making these tradeoffs by seeing to it that prices reflect costs. Great emphasis is given to how the invisible hand of the market can direct individual choices to socially correct prices, assuming appropriate conditions.

Having presented the "right price" vision, the texts present a discussion of those conditions that can lead to a distinction between social costs and private costs due to externalities. Much of students' training involves developing an understanding of the attributes of those "right" sets of prices, and the problems that can prevent those right sets of prices being reached.

This neoclassical vision is embodied in a set of models and concepts. In micro, these models and concepts include: the production possibility frontier and the concept of comparative advantage; supply demand analysis; rational decision making, as inherent in indifference curve analysis; the theory of the firm, and marginal cost analysis; the model of the competitive firm and monopoly; and social welfare analysis, as embodied in consumer and producer surplus. Each of these models and concepts contributes to instilling the neoclassical vision into students.

While the neoclassical micro vision is clear, the neoclassical macro vision is hazy. Until the 1980s it involved a vision of market failure on the aggregate level, with the need for government to correct that market failure with appropriate macro policies. That vision was built into the texts with the Keynesian AE/AP model at the principles level, and the IS/LM model at the intermediate level. In the 1980s that Keynesian vision has changed into a vision more consistent with the neoclassical micro vision. Aggregate markets work, except for externalities, and while there may be short-run macro externalities, in the long run the market will guide the economy to its natural rate of output and its long-run optimal growth rate.

This change in vision has caused confusion in the macro texts, and has necessitated a change in the models to make them consistent with the changed vision. Thus, we have seen the recent introduction of the AS/AD model into the principles texts, and a recent attempt by Mankiw to downplay even this model and switch the focus to long run growth. In intermediate books we have seen the Solow growth model replacing IS/LM as the first model presented to students. With the shift in emphasis and rise in importance of neoclassical growth theory, macro is being pulled into the same neoclassical "right price" vision as micro.

Many economists of varying persuasions, and at various levels, do not share this textbook vision. In highly technical work, and in their informal discussions among friends, many economists recognize that one must go far beyond the neoclassical vision of the texts to understand the economy. In discussions among themselves they are quite willing to recognize, and even make fun of, the simplicity of that textbook neoclassical vision. However, when it comes to teaching, those same economists will often teach unquestionably the neoclassical vision, justifying what they do on pedagogical grounds. They argue that to present students with the problems of the mantra simultaneously with the mantra undermines the pedagogy.

This approach to teaching the principles course has a significant effect on the composition of the economics profession. Those who find the neoclassical vision discordant with their understanding of economic reality drop out of economics, while those who find it resonant go on. In this manner the neoclassical vision becomes self-reinforcing, and it precludes other visions from replacing it.

The Santa Fe vision is quite different than the neoclassical vision; it is one of emergence and process, not equilibria states. It is a vision of interdependent organizations and hierarchies that must be taken into account in any analysis. It is a vision in which there is no "right price," and no definitive answers about whether the current market structure is the best we can do. Rationality is inductive rather than deductive, and the support for the market as an organizing structure comes from our lack of knowledge of social processes, rather than from our theoretical proofs that markets are the best way to organize an economy. It is a vision of continual adaptation at various levels, and of perpetual novelty. Within the Santa Fe vision, tradeoffs are much more complicated than those presented in the neoclassical vision, and the integration of economics with other fields becomes central to the analysis.

Calling the Santa Fe vision a complexity vision is misleading. Both the neoclassical and the Santa Fe approach recognize complexity; they also both assume that underlying complexity is a simplicity. The difference between the two views is that they find this simplicity in different places. In the neoclassical model, the simplicity is to be found in the static structure of system; production functions are homogeneous, and choice sets are susceptible to marginal analysis. In the Santa Fe vision the simplicity is not to be found in the comparative static structure of the models or in linear dynamics. In the Santa Fe approach, the simplicity is to be found in the rules underlying dynamics. It is to be found in the simple iterative dynamical laws operating within non-linear dynamical systems. In the Santa Fe vision one's understanding of a phenomenon is not of the phenomenon directly but of the process that led to that phenomenon.

There are quite different limits to our understanding in the Santa Fe model and the neoclassical model. In the Santa Fe model one cannot analytically understand the system over all time, and, even if one fully understood the underlying analytics, one cannot perfectly predict the future. The issue of economics is not one of getting the prices right, but of understanding the varying roles that prices are playing. Given a process many alternative equilibria are possible.

The Santa Fe vision is not a new vision. Variations of it have been around for a long time in economics. For example, Adam Smith, Frederick Hayek, and J.M. Keynes all had aspects of the Santa Fe vision in their analyses. Such views, however, were downplayed by textbooks, and their contributions often forgotten. Somehow, whenever their ideas were translated into formal models, that Santa Fe vision was eliminated. The reason is that the models presented are all based on an assumption of static simplicity. Within these static models there is no room of Santa Fe simplicity and the view that complex phenomena can result from very simple dynamic laws and dynamical systems. That Santa Fe simplicity is to be found in the competitive process, not in the competitive state.

These forces pushing pedagogy away from competitive process ideas have existed for a long time, and, if nothing had changed, I would believe that there was little chance of integrating the Santa Fe vision into the texts. But in the 1990s something has changed. What has changed is the possibility of formally "solving" the dynamic model, or at least coming to tentative conclusions based on the model, within the Santa Fe vision. The second volume of the Santa Fe institute provides a glimpse of what some of those conclusions are. With computer simulations, insights into an increasing array of dynamic processes can be seen.

As this work progresses, and becomes better known, more and more professors will be amenable to introducing Santa Fe ideas into their courses. Instead of being seen as atheoretic ideas, such ideas can be seen as consistent with the latest developments in theory. Thus, used carefully, they can modernize a book. Programs embodying these solutions can be easily used and integrated into a principles course. (Here a survey paper by Jeff on computer programs and videos that introduce the Santa Fe vision might be relevant.)

The focus of these suggested ideas will be on the broadening of the issues as presented in the principles text so that they better fit the real world. Thus, the core of what is currently taught will remain, but it will not be presented as the entire analysis. It will be supplemented with concepts, terminology ,and models more compatible with the Santa Fe view.

The hurdle to overcome in doing this is that most economic professors are satisfied with presenting the limited model. They believe it provides a reasonable description of part of the decision-making process, and gives students a model, or at least a conceptual picture, that they can understand. For professors to want to include the Santa Fe vision, they must believe both that it is adding something to students' understanding and that it is not making the analysis too difficult to understand. Introducing Santa Fe ideas into the texts as an extension of existing ideas, rather than as a replacement for them, is the most likely way of meeting both of these requirements.

Let me now turn to three conceptual ideas that I believe will make the texts more Santa Fe-friendly. After briefly discussing these ideas, I discuss some of the amplifications of models that might be made to include these ideas in the texts.

Current economics texts spend little time discussing induction and deduction; they simply present economics as deductive analysis. A Santa Fe-compatible book would explore the induction/deduction distinction more carefully. It would emphasize how knowledge is based upon observation, and how those observations are used in combination with deduction to arrive at a decision. It would explain how economics, and every science, has always involved a balancing of induction and deduction.

Translating that balance down to the textbooks is difficult, as textbook models generally embody previous economists’ inductive insights that have then been translated into models. Those models are taught deductively. For example, the supply/demand model embodies numerous insights about the world that previous economists have gathered from experience. Students are not taught to gather that same information, or to inductively determine a model. Instead, they are taught to go through a variety of exercises that deductively flow from an understanding of the supply/demand model.

This focus on deduction rather than induction also shows up in the textbook discussion of rationality. That discussion is closely tied to marginal utility or indifference curve analysis. With this analysis students are taught to differentiate between the income effect and the substitution effect, and are taught to think of standard optimality conditions when thinking about choice. If the economy were a two-good economy, that deductive view of choice might be sufficient, but for the broad set of choices individuals make, it is insufficient. Thus, to integrate the Santa Fe approach into the texts, it is necessary to broaden that approach–to ask students how they would decide when faced with more complicated issues.

One way to broaden it is to present some real-world choices and ask students how they made those choices. In other words: Direct students toward the inductive approach, rather than simply presenting the deductive approach. It will quickly become apparent that many of the choices students make are complicated, and do not fit the simple choice model easily. One can then ask students whether it is reasonable to simplify the complicated analysis that is to be represented by the deductive choice framework. In many cases, I suspect that students will believe that it is not. For many decisions it will be obvious that the "rational decision" will be too costly and too time-consuming. Instead, for some decisions, individuals will rely on the others to guide them as a way of saving on decision-making costs. By integrating costly decision-making into the texts, one will have a framework for integrating an analysis of fads and speculative bubbles. The "costly decision" framework also can be used to develop "rules of thumb" decision-making, creating a natural introduction to concepts such as attractors and focal points. Once these concepts are used, students who are interested in them will have an incentive to develop the technical foundation for them; they will be led to Santa Fe work.

Having developed the need for a broader-based choice theory, the text can draw a distinction between "simple choice" and "complex choice." The analysis of simple choice would be very similar to what we currently have in the texts. But students would also be given a discussion of how more complicated choices are generally made. Issues in cognitive psychology could be introduced, and a simple example can be given, such as Brian Arthur's "go to the café" example. Such discussions will lead naturally to an analysis of advertising, and the approaches advertisers use to make their good a focal point. Thus, this "costly choice" extension can be used to explain questions such as: Why is Michael Jordan paid $40 million a year to endorse products? Why does Nike spend millions on placing its logo on sports stars, and on "just do it" advertisements? and Why do localities pay so much to attract a ball team to their city? Complex choice theory can explain such phenomena; simple choice theory cannot. (Stodder's paper will explore related issues. )

In the texts, choices are invariably presented as marginal choices. In the real world, marginal choice analysis is generally insufficient. Many real-world decisions are not marginal at all; they are sequential, and non-marginal, or all-or-nothing choices. For students to have an appreciation of the Santa Fe vision, they need an understanding of such choices. Again, this understanding can be conveyed to students by inductively leading them through some of the decisions they have made. Consider, for example, a student's decision to take an economics class. That decision is a sequential decision that is the end product of a long string of choices. In high school the student made a decision about how much to study. Then she made a decision about which school to go to and what major to take. It is possible that she signed up for another course, only to find that the teacher was not to her liking, causing her to shuffle courses to find a replacement at the right time. The end result may be the choice of taking economics, but the choice itself can only be understood from the sequence of decisions that led her to that ultimate choice. At each point along the way, the relevant costs changed; some choices were irreversible, others were not. The same issues hold in production theory as Herbert Scarf (JEP) points out. To meaningfully analyze real world decisions, non-marginal and sequential analysis is an important element. (Barkely's paper will explore how the math we teach would change to better fit the Santa Fe vision.)

In making sequential decisions the relevant costs are often changing, and, often, the options considered will not be marginal decisions, but non-marginal. In such cases the only choice criterion one has is: Total relevant costs must be below total relevant benefits. Given that criterion, there are a number of sub-criteria relevant for special cases. The current texts treat the marginal choice criterion as if it were the only case.

One of the major differences between the Santa Fe vision and the neoclassical vision concerns equilibrium. The neoclassical vision is centered around static equilibrium; the Santa Fe vision is centered around dynamic processes of an economy far from a static equilibrium. The Santa Fe approach focuses on the competitive process, not the state of competitive equilibrium. In it the competitive state is something that is almost never even closely approached; long before an industry moves to a competitive state, a technological change will change the industry, creating a further disequilibrium. Competition often comes from without, not within, and involves changes in products. In most texts this view is lost because products are unchanging. If students are to be open to the Santa Fe approach, they need some sense of the enormous change in products that occurs, and how successful firms are continually redefining themselves, both as what products to produce and how the operate.

Given the pervasiveness of the equilibrium concept, it seems unlikely that the current concept of equilibrium will change. The concept of equilibrium is so deeply engrained in the pedagogy of economics that the Santa Fe view that interesting dynamical systems, such as the economy, operate far from equilibrium has little place in current texts, and it seems unlikely

In thinking about this problem I was reminded of a story my 6-year-old son was recently assigned to read. It is a tale of an elf who is caught, and forced by his captor to tell where his gold is buried. The captor has to go get a shovel to dig up the gold. He makes the elf tie a ribbon around the tree, and promise not to remove it, so that when the captor comes back with the shovel he can identify the tree under which the treasure is buried. The elf keeps his promise, but upon the captor's return, the gold is unfindable because the elf has tied a ribbon around every tree in the forest. Essentially, this same trick can help introduce the Santa Fe vision of competitive process into texts. The concept of long-run equilibrium will not be so confining if there is not a unique long-run equilibrium. If the textbooks, early on, introduce the concept of multiple equilibria, students will recognize the importance of institutions and conventions in economics in choosing among those equilibria. Discussion of moving from one equilibrium to another can be used to introduce competitive process ideas and how the competitive process operates within institutions, and changes those institutions.

A standard example of multiple equilibria involves society's choice between driving on the right side, and driving on the left side. Two equilibria are possible. Britain and Japan have chosen one; the U.S. and Germany another. Each choice is equally efficient, and as long as all the people while in a particular country choose the same side, there is no problem. But one side or the other must have been chosen by the country. As products evolve, society is continually making such choices, and enormous resources are spent influencing those choices. It is a dimension of competition that current textbooks miss completely.

Multiple equilibrium allows an introduction of the problem of industry standards and a number of real-world examples, such as the Beta/VHS fight, can be highlighted in case studies. When there are multiple equilibria, firms will fight for growth and market share, and often will accept temporary losses, if by doing so they can direct the economy to an equilibrium that benefits them in the long run. Those fights are a good introduction into the Santa Fe vision.

Having introduced the concept of multiple equilibria, one can discuss the concept of equilibrium selection mechanisms. It is here that brief discussions of evolutionary game theory and simple repeated games can be used to firm up students' understanding of such processes. Computer simulations can also be developed that consider the solution to such multiple equilibria games, and that show the importance of initial conditions, and initial positioning, and institutional structure.

The three ideas discussed above can be integrated into text discussions in a variety of ways. In many ways, doing so would liven up the books, better relating the ideas in the texts to real-world events. For example, the competitive process is an idea that students would like. Many more real-world stories can be told about the competitive process than can be told about the competitive state. There is, however, the problem discussed above. That problem is that the dynamic concepts don't neatly fit in with the static models. While students like discussions of the real world when they are reading for fun, when they are studying for exams they lose their interest in the real world and want the texts to cover the material that will appear on their exams. Inevitably, what primarily shows up on exams, especially the multiple choice exams that predominate in introductory courses, are questions about models. Ideas that make it harder to pull definitive conclusions from models are distracting ideas, and eventually will be culled from the texts.

Thus, to integrate these ideas into the textbook canon, one must either integrate these ideas into the core models that are taught, or develop compatible models that can modify the neoclassical mantra to one more conducive to the Santa Fe vision. I now turn to a discussion of how three of the canonical models might be modified to fit the Santa Fe vision.

The first graph students are generally introduced to is the productions possibility curve. Through this graph students are introduced to opportunity costs–the infamous guns/butter tradeoff. For the Santa Fe approach to be integrated into this model additional aspects of decisions need to be added to the models: sequential decision making, increasing returns, non-linearities, and learning by doing. Current texts avoid the issues by presenting the production possibility curve as a concave curve, with that shape justified by an assumption of diminishing marginal returns.

To integrate Santa Fe ideas into what is presented, the production possibility frontier must be presented (1) as a more complicated curve–one that is not concave over its whole range, and (2) as a constantly evolving curve--one which changes as decisions are made. To integrate the first idea, the text might present two ranges of the curve. The first would be the standard range where diminishing returns predominates. The second would be a range where increasing returns predominates, making the production possibility curve convex. Within this convex range, there are gains to specialization. Tradeoffs still exist, but those tradeoffs will not be marginal. In the convex range if one follows the normal marginal conditions one will be minimizing, not maximizing. These gains to specialization were very much a part of Adam Smith's view of the working of markets, but this role of specialization has disappeared from the texts.

Sequential decision-making can be presented by presenting a decision tree along with the production possibility curve. The text would explain that many decisions are costly to reverse, and that, when such costly-to-reverse decisions are made, the choices facing the decision-maker change. Decisions made in a costly reversibility context are fundamentally different than decisions made in a costlessly reversible context. The pouring of concrete, the sawing of a board, the cutting of a diamond, the choice of schools by students–all these can be presented as examples. Within a decision tree, sequential decision, framework, the tradeoffs are continually changing. Instead of decisions being presented in a static one-time-decision framework, the decision tree leads students to picture decision making within an ongoing decision framework.

One of the most important models that can help integrate the Santa Fe vision into the texts is the experience curve, also called the learning curve. This curve (to be discussed by Rothchild) is well known in business but is seldom discussed in economics texts. The reason is that it is a dynamic concept–costs decrease over time, not necessarily because of investment the firm has made, but because of the experience it has gained. Simply producing is a type of investment. Acceptance of such a curve undermines any simple concept of static equilibrium, and creates a tendency toward monopoly–a tendency that is quite incompatible with the standard textbook view of competition.

The learning by doing that accounts for the experience curve is quite different than increasing returns to scale that accounts for a downward sloping long-run cost curve. The first is a dynamic concept; the second is a timeless concept. The experience curve has nothing necessarily to do with changing the scale of production facilities; it has to do with the volume. By producing in the short run one gains a comparative advantage in the long run. Thus, when the experience curve is operative, comparative advantage is endogenous, and there is a natural tendency toward expansion and growth of those who are producing within the economic system. In the long run, the experience curve causes competition to lead to monopoly, not to a competitive state of multiple producers. In the Santa Fe vision that monopoly is prevented by the development of competition from without, such as the development of new products, which undermine that monopoly. Santa Fe competition involves continually new products and technological change, and that technological change offsets the natural tendency toward monopoly. Thus, rather than being exogenous, technological change is fully endogenous to the competitive process.

The nature of competition in an industry depends on where the economy is on its experience curve. Two types of industries might usefully be distinguished: technologically dynamic industries and technologically stagnant industries. Technologically dynamic industries are industries on the steep portion of the learning curve–the computer industry is an example. They have enormous cost reductions over time. Technologically stagnate industries are industries at the tail of their learning curve–the refrigerator industry might be an example. Their costs are relatively constant over time, and fluctuate primarily with input prices.

Current texts present all industries as technologically stagnant. Presenting the experience curve early on in microeconomic texts part of the course will direct students to think in terms of technologically dynamic industries. Such industries will be characterized by enormous competition, new entrants, and a search of an industry standard. Once a standard becomes apparent, the nature of the competition changes, and some type of oligopoly market structure develops.

Rich, real world stories can be told around the experience curve, incorporating the decision tree analysis above. The lesson students learn from these stories is that in a dynamic framework the concept of "relevant costs" is not a fixed or precise concept that can be determined from accounting data alone. Determining relevant costs requires judgment, expectations, and enormous institutional and technical knowledge. Such knowledge often is unavailable. Thus, the learning curve sets up a framework of economics within a broader uncertainty context that better helps students relate what they are learning with the real world around them.

All discussions of the learning curve that I have seen have centered on individual industries. But since the aggregate economy is a complex agglomeration of industries, there should be a corresponding aggregate learning curve for the entire economy. This aggregate learning curve relates aggregate output with the growth rate in long-run productivity and thus provides a bridge between the long run and the short run in the same way that it did in the industry case. If there is an aggregate experience curve, long-run growth depends on short-run expansion, and it is impossible to separate out the two. Thus my latest attempt to introduce Santa Fe ideas into macro involves the aggregate learning curve as an important determinant of growth and an explanation of how aggregate economies can be on a different position on the learning curve. Movement along an aggregate learning curve makes technological waves of growth possible.

The multiplier model is currently out of favor as focus has shifted from the short run to the long run. From a pedagogical standpoint this has caused professors grief because the multiplier model was a wonderful crutch that provided a core model for the macro that was taught through the 1980s. With the demise of the multiplier model, there has been a demise of Keynesian economics. But with learning by doing and the experience curve, the multiplier model can be reintegrated into the theory of long run equilibrium as a simple example of path dependency. To do so one would need to have some autonomous production dependent on autonomous expenditures (creating something like the multiplier-accelerator model), but having provided that, one can develop a richer model that is open to Santa Fe ideas. (Bob Prash will discuss such issues.)

As should be clear from the discussion in this paper, I do not see the process of integrating Santa Fe ideas into the texts as easy. New ideas do not find fertile fields in entrenched professions such as the economics teaching profession. But to the degree that we succeed in making the texts more Santa Fe-friendly, we will make it easier for deeper Santa Fe ideas to make their way into the profession. Thus I think it is an important undertaking that deserves serious consideration.